OpenAI Launches Lockdown Mode Against Prompt Injection Attacks

OpenAI has unveiled a Lockdown Mode for ChatGPT designed to prevent sensitive data from being exfiltrated through prompt injection attacks, where malicious instructions hidden in content manipulate the AI into leaking information. The feature does not fully eliminate the vulnerability but is aimed at reducing the likelihood that private data gets exposed during an attack. It arrives as AI agents handle increasingly sensitive enterprise workflows.